Emotion mining research on micro-blog

In order to identify the emotion expressed in corpus and to estimate the feelings conveyed by micro-blog data, in this paper, we categorize emotional words, build attitudinal words weight dictionary (WD) which consists of 1342 words, and construct self-defined negative words dictionary (NWD), degree words dictionary (DWD) and interjection words dictionary (IWD). We then process classified statistics on 2213 micro-blog items, finding that items whose first sentence express the main idea account for 23.8% of all the complex sentences, and items whose last sentence express the main idea 51.3%, respectively. We first calculate the weights of each clause in a micro-blog item, then take special treatment to the first and last sentence, finally we add up all the weights to get the emotional index (EI) of the item. Test results from the micro-blog emotion weight calculator (MBEWC) developed in C# are cross-checked and reach an accuracy rate of 80.6%.